Testing Homogeneity of Variances

This note is intended to acquaint the reader with a useful hypothesis test which apparently has gone unnoticed by the "statistics consumer." This test was devised and studied by Levene (1960) and is appropriate for testing the null hypothesis, H0, that J samples come from populations with the same variance, σ. Three instances in which testing for heterogeneity of population variances is worthwhile come to mind: a) when one wishes to make inferences about population variances because they are of scientific interest, b) when one suspects heterogeneity of variances in an analysis of variance in which not all factors have fixed effects, c) when one suspects heterogeneity of variances in a fixed-effects analysis of variance in which the numbers of observations in the groups are widely disparate.